AI Deployment Risk · Governance · Pre-Adoption Diligence

I wrote the book on when AI deployments fail.

"Hire me so yours doesn't."

Independent AI risk assessment and governance consulting for small professional firms, rooted in a documented six-risk framework and hands-on evaluation of frontier models.

When AI Is Wrong for the Job by Zachary A. Perlman — a practical framework for leaders, builders, and decision makers
When AI Is Wrong for the Job
Zachary A. Perlman · Published June 2026
View on Amazon →
Methodology
Six-Risk
Framework
Documented across real deployment failures. Not proprietary jargon.
Certification
AIGP · IAPP
AI Governance Professional, credentialed 2026.
Independence
No Vendor
Affiliations
No referral fees. No preferred tools. The recommendation is the work.
Pricing
Fixed-Fee
One price. No retainer. No billable-hour surprises.
01 — Methodology

The Six-Risk Framework

The methodology behind the book, a systematic lens for evaluating AI deployments before and after they go live. Every engagement is run against these six categories. No proprietary black box; you see exactly where the risks land and what to do about them.

Risk 01

Data & Confidentiality

What client data enters the model, where it goes, and who can see it downstream. The Samsung leak was this category.

Risk 02

Output Reliability

Hallucination rates, citation fabrication, and what "close enough" actually costs in a professional context.

Risk 03

Verification Tax

The hidden time cost of checking AI output. Whether the "hour saved" is real or a liability in disguise.

Risk 04

Liability Exposure

Who signs off on AI-assisted work and what professional liability rules apply when it goes wrong.

Risk 05

Vendor Accountability

What the terms of service actually say and what recourse you have when the tool changes, fails, or disappears.

Risk 06

Scope Creep

How a narrowly-scoped pilot expands into unsupervised use across the firm, and how to prevent it structurally.

02 — Services

One engagement. Five deliverables.

The starter package is designed for small professional firms (solo to ten-person practices) already using AI or actively evaluating it. Fixed fee. Concrete outputs. No retainer required to start.

Starter Package

AI Risk Assessment

$2,500–5,000 fixed fee
  • AI-use inventory of the firm
  • Risk memo against the six categories
  • Written AI usage policy
  • Vendor due-diligence assessment
  • One staff training session
Advisory

Pre-Adoption Diligence

Custom scoped per engagement
  • Evaluation of a specific tool or platform
  • Four-move memo for leadership
  • Scoped-use policy for the application
  • Incident log template
  • Cost-benefit worksheet
Expanded

Ongoing Governance

$5,000–10,000 multi-attorney firms
  • Full governance framework design
  • ISO 42001 gap assessment
  • Regulatory readiness review
  • Quarterly risk update memos
  • Staff training, multi-session
03 — Law Firms

The stakes are higher when you bill by the hour.

Small law firms face AI risks that differ in kind from corporate deployments. Confidentiality rules don't bend. Professional liability follows the attorney who signed off. And the tools being adopted were not built with your obligations in mind.

Chapter 14 of the book maps ABA Opinion 512 and the Model Rules directly to AI use in legal practice. The consulting engagements apply that research to your specific firm.

ABA Opinion 512: competence, supervision, and confidentiality requirements in plain English
Data What happens when client trust documents enter a third-party AI model
Liability The sanctions enforcement wave and why it lands on whoever signs off
Vendors How to vet an AI vendor when you're a three-person firm
Cost The verification tax: what the "hour saved" really costs a billing practice
Why this niche

Estate attorneys, title companies, and solo practitioners are adopting AI tools at pace, often without a compliance officer, a technology committee, or any process for evaluating what they're putting client data into.

The same diligence the book teaches employees to demand can be commissioned from the top. Same framework. Different table. You get an independent assessment with no vendor stake in the outcome.

04 — About

The credential is the work.

The consulting practice is built on documented research, not theory. When AI Is Wrong for the Job is a systematic analysis of real AI deployment failures, mapped against a repeatable risk framework and written for leaders, builders, and decision-makers who need to act on what they find.

That research is what gets applied to your firm. Conservative estimates. Honest citations. No overclaiming. The same credibility standard the book holds itself to.

Book When AI Is Wrong for the Job (2026). Documented analysis of AI failure modes in professional deployment contexts.
AIGP AI Governance Professional certification, IAPP, 2026
Eval Paid adversarial evaluation of frontier models: red-teaming and safety-track work across multiple platforms
Writing 20+ years professional writing; published technical security writing (The Hardened Mesh)
Site caseagainstai.org, public-resource edition of the framework and ongoing research archive
Incidents in the framework
Samsung Electronics

Engineers pasted proprietary source code into ChatGPT. Immediate confidentiality breach with no recovery mechanism and no contractual recourse.

Air Canada

Chatbot gave a customer incorrect bereavement fare policy. Tribunal held Air Canada liable for its AI's statements regardless of disclaimers.

Sanctions Enforcement Wave

AI-assisted compliance tools missed flagged entities. Liability fell on the firm that relied on the output without independent verification.

Workday / iTutorGroup

AI screening tools introduced discriminatory bias in hiring decisions. EEOC scrutiny and litigation exposure followed the employer, not the vendor.

05 — Writing

Applied research for practitioners.

Ongoing articles applying the book's framework to specific professional contexts, published here and on LinkedIn. Written for the firms this work is designed to serve.

What happens when your paralegal pastes a client's trust into ChatGPT

A step-by-step breakdown of the data flow, the confidentiality exposure, and what Rule 1.6 actually requires of supervising attorneys.

Coming soon

The sanctions wave and why it lands on whoever signs

How AI-assisted compliance tools are failing at the margins, and the enforcement pattern that's emerging around them.

Coming soon

ABA Opinion 512 in plain English for small firms

What the opinion actually says about competence, supervision, and confidentiality, and what a three-person estate practice should do about it.

Coming soon

How to vet an AI vendor when you're a three-person firm

The due-diligence checklist from the book's appendix applied to a real vendor evaluation: what questions to ask and what answers to reject.

Coming soon

The verification tax: what the "hour saved" really costs a billing practice

Breaking down the hidden time cost of reviewing AI output, and the math that changes the ROI calculation for attorneys.

Coming soon

An AI usage policy for a solo estate practice

A worked example of the scoped-use template from the appendix, adapted for a solo practitioner with two paralegals using three AI tools.

Coming soon

Find out where your firm's AI exposure actually is.

Fixed-fee engagement. No retainer.
No vendor stake in the recommendation.

Schedule a Conversation